A Traffic Monitoring Method Using Accumulative Difference Images
Kornchanok Krajangyao, Toshiaki Kondo, Waree Kongprawechnon, Jessada Karnjana, Atsushi Takahashi · 2024
This paper presents a traffic monitoring method to count the number of vehicles from video surveillance on a highway. The proposed method uses the accumulative difference images (ADI) for extracting the vehicles from stationary background scenes. The result of ADI is a sequence of images, namely, x-y planes. However, we cannot count the number of vehicles in full frame directly because the vehicles far from the video camera are small and occluded. The horizontal-temporal cross-section plane is applied for ADI result that is the 3-D data (x, y, t) to slice into y-t plane at a certain level of x to observe the moving object or vehicles on video within one frame. To count the number of vehicles in cross-section planes within 1-D projection, we propose to separate the lanes of the road to avoid missing vehicles caused by connection of vehicles that extracted from original video. The result shows that counting the number of vehicles with full frame on x-y planes is less accurate by the video environment than we propose to use horizontal-temporal cross-section method to observe the y-t planes.